Object-oriented video tracking originates from the eighties of the twenties century. The purpose of motive target tracking is to split out the essential information and detect and tracking the targets. It is very important technology in image processing and computer vision area. And it has extensive applications in robot navigation, intelligent monitoring, medical image analysis, video compression and transmission.There're abundant raw data in video media. The strong correlation between adjacent frames and the time-varying dynamic modes in time domain make it possible in detection, segmentation and recognition. The motive target detection means abstracting the target information from the video, while the target tracking means continuous detection for the target and specifying its motion path.This article proposed a motive object detection arithmetic based on background statistics model. The arithmetic is set on the basis of subtraction between frames method and W4 model proposed by Hmitaoglu and Harwood of Maryland University for its model construction, and is set on background subtraction method for model updating, and jointly gave out a self-adaptive background-subtraction object detection arithmetic. The arithmetic has very good robust for light change, at the same time, it fully used the information of the foreground; it can construct its background model in the first frames quickly, and improve the correction rate of foreground detection.For the motive object tracking, the article analyzed the shortcomings of Mean Shift theory, corresponding to the lack of capacity for tracking the fast-moving objects and the occlusion, it proposed a improved method and gave out its detailed description. Its main method is jointly considering the key character region detected by the detection arithmetic and Mean Shift method, and then uses the logic space relationship to determine the location of the object being tracked quickly. The improved algorithm can track the fast-moving target well and alsoThe method above is tested with Visual Studio 2005 in Windows operating system based on Intel X86 platform. Seeing from the result, the method I proposed is feasible, and it can basically meet the demands for real-time tracking. At the same time it has better robust for occlusion compared to normal tracking arithmetic.